Content
56%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The content is highly actionable for the core tasks (API usage, provider configuration, custom provider registration) with executable code and specific file-level rules. Its main weakness is scale: a monolithic 545-line file inlines deep operational detail (consent, noise filtering, event schemas) that belongs in reference files, costing token efficiency and progressive disclosure.
Suggestions
Move the Error Capture, Cookie consent, MCP Server Events, and Default Baseline Events sections into separate reference files (e.g. references/error-capture.md, references/consent.md) and keep one-line pointers with "See X when..." triggers in SKILL.md.
Tighten the long bullets (e.g. "Browser outcomes are bounded, too", "Rates are events, not log lines") into short rule statements plus property tables, cutting run-on edge-case narration.
Lead with a short Quick Start (set one env var, call track()) before the API section so the common path is reachable without reading the full document.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body runs ~545 lines of dense edge-case prose, e.g. the enormous "Browser outcomes are bounded, too" bullet and the multi-paragraph cookie-consent section detailing sampling rates, cookie attributes, and Consent Mode mechanics. It never explains generic concepts Claude already knows (so not 1), but the sheer volume of inlined detail is noticeably verbose even though accurate. | 2 / 5 |
Actionability | Executable, import-bearing code for track(), identify(), and registerTrackingProvider(), env-var tables with defaults, and concrete rules like "update both server/posthog-config.ts and the mirrored worker emitter in deploy/build.ts". Below 5 because large sections (baseline event property catalogs, MCP event tables) are descriptive schemas rather than copy-paste-ready guidance. | 4 / 5 |
Workflow Clarity | "How It Works" gives a clear 4-step startup-to-fan-out sequence, and verification hooks exist ("posthog-config.spec.ts pins the two outputs together", "check what an event looks like in its UI, not just that the request returned 200"). Below 5 because mandated multi-file edits lack an explicit validate-and-retry loop; the destructive/batch cap does not apply since tracking is non-destructive. | 4 / 5 |
Progressive Disclosure | No bundle files exist; ~400 lines of error-capture, cookie-consent, MCP-event, and baseline-event detail that clearly belongs in separate reference files are inlined in SKILL.md. Scored above 2 because the file does have consistent, well-named section headers throughout, giving real structure — but nothing is split out or referenced one level deep. | 3 / 5 |
Total | 13 / 20 Passed |